This study uses electroencephalography (EEG) data to construct an emotion identification system utilizing a deep learning model. Modeling numerous data inputs from many sources, such as physiological signals, environmental data and video clips has become more important in the field of emotion detection. A variety of classic machine learning methods have been used to capture the richness of multimodal data at the sensor and feature levels for the categorization of human emotion. The proposed framework is constructed by combining the multi-channel EEG signals' frequency domain, spatial properties, and frequency band parameters. The CapsNet model is then used to identify emotional states based on the input given in the first stage of the proposed work. It has been shown that the suggested technique outperforms the most commonly used models in the DEAP dataset for the analysis of emotion through output of EEG signal, functional and visual inputs. The model's efficiency is determined by looking at its performance indicators.